摘要
传统的轨道扣件检测主要依靠二维图像,存在检测精度低等问题。因此,提出一种基于2D与3D激光图像的轨道扣件状态智能检测方法。通过图像灰度化与线性加权融合处理构建三组轨道扣件数据库;利用YOLOv5进行轨道扣件状态的自动检测与螺栓区域定位;提出基于区域收敛的螺栓分类法,区分道钉与螺母区域;结合三维深度信息并确定阈值,实现螺栓松动检测。实验结果表明:经过线性加权融合处理的模型精确率比另两组实验高3.2%、11.3%;同时,提出的检测方法能够实现轨道扣件状态智能检测及螺栓松动的自动检测,具有较强的适用性。
The traditional rail fastener detection mainly relies on 2 D images, which has problems such as low detection accuracy.Therefore, an intelligent detection method for the state of rail fasteners based on 2 D and 3 D laser images was proposed.Three groups of rail fastener databases were constructed by image grayscale and linear weighted fusion processing.YOLOv5 was used to automatically detect the state of rail fasteners and locate the bolt areas, a bolt classification method based on regional convergence was proposed to distinguish between spikes and nut areas, the bolt loosening detection was realized by combining the 3 D depth information and determining the threshold.The experimental results show that the accuracy of the model processed by linear weighted fusion is 3.2% and 11.3% higher than the other two groups of experiments.The proposed detection method can realize intelligent detection of track fastener status and automatic detection of bolt loosening, which has strong applicability.
作者
陈文婷
罗文婷
李林
秦勇
温王鹏
吴镇涛
CHEN Wen-ting;LUO Wen-ting;LI Lin;QIN Yong;WEN Wang-peng;WU Zhen-tao(College of Transportation and Civil Engineering,Fujian Agriculture and Forestry University,Fuzhou 350108,China;College of Transportation Engineering,Nanjing Tech University,Nanjing 210009,China;State Key Laboratory of Rail Traffic Control and Safety,Beijing Jiaotong University,Beijing 100084,China)
出处
《仪表技术与传感器》
CSCD
北大核心
2022年第11期88-95,共8页
Instrument Technique and Sensor
基金
国家重点研发计划项目(2021YFB3202901)
福建省高校产学合作重大项目(2020H6009)。